CN1303497A - Distributed computing environment using real-time scheduling logic and time deterministic architecture - Google Patents

Abstract

A system is presented for a plurality of computer nodes in mutual data communication to execute logic in a coordinated manner. The approach features aligning the definition of the time in the nodes so that each node has essentially the same definition of time,defining a schedule of communications for each computer node, with the schedule having future moments defined for initiating each communication, communicating each schedule to its respective node, and initiating each communication at the time defined in the schedule. The computer network has a message carrier supporting multiple communication paths; a plurality of computers coupled to the message carrier, a clock system for maintaining time synchronism among the computers; and a scheduler coupled to the message carrier for determining the sequence of communication between the computers. The message carrier establishes communication between selected computers based on the time-deterministic tasks being performed by the computers. A computer node in the plurality of computer nodes executes logic in coordination with the other nodes by defining a communication request list of each communication needed by the computer node with any other computer node respective to a future time period, sending the communication request list to the master scheduler, receiving back a communication schedule back from the master scheduler, defining a local task schedule from the communication schedule, and then executing the communication at the scheduled time.

Description

Adopt Real-Time Scheduling logic and time to determine the DCE of structure

The present invention relates to data processing and control system field, particularly a plurality of computers and in network, carry out the integrated system of their logic with cooperative mode. Although the described system of this specification is also useful to the effort that relates to multicomputer processor, parallel computer processor and network computer processor, it is used at first is the computer control of the equipment that adopts for large entity scale facility (for example chemical industry facility).

When computation requests or geographic restrictions need to be calculated facility physics and scatter (for example being used for process control large, the upper chemical plant that disperses of geography), central authorities' coordination of employing calculation task may be of value to practical factor (such as the product quality). The invention describes a kind of system and method, can make geographical upper many computers that disperse become the network of the internal act that can realize time coordination. This joint network can make task coordinate and can realize that other is of value to the character of the performed task of native system,, safety reliable such as redundancy and high flux.

5,428, No. 769 United States Patent (USP)s that announce June 27 nineteen ninety-five have been authorized Glaser, Robert S; Hoy, Robert S; Fernandez, G.Paulh and Grai, Timothy J, name is called " the process control interface system with unit, triplex level redundancy far field ", and it has been described one and has illustrated that this class is for the embodiment of the system of unit, the far field enforcement real-time interconnection of the series of physical distribution of process control system. Although this networking method can be each system's addressing effectively within each time cycle when needed, thereby this system is applicable to multiple process control situation; But, if the variable of all exchanges all is stable, have in these addressing circulations so many all be unnecessary-from the viewpoint of network efficiency, only just provide the scheme of communication more succinct inputing or outputing in the situation that needs to change.

The physics of computing equipment scatters that to coordinate to combine in the high-performance processor with behavior feature also with the central authorities of calculation task also be a main concept. Show (Morgen Kaufmann at John L.Hennessy and David A.Patterson; San Mateo, California; 1990; The sort of vector processor of describing in 7.1 and 7.2 joints (" vector processor ") of " Computer Architecture; a kind of quantitative approach " the 351-363 page or leaf) is exactly a proof of this system, and E.F.Gehringer, D.P.Siewiore and Z. Segall are at " parallel processing-Cm^*Experience " (Digital publishing house is through Editorial company, Rockport, Massachusetts; 1987) the sort of Cm from the 11st page to the 18th page of discussion^*Design provides second example of these class methods.

Coordinate also to have used such as J. William Gavett at " produce and operate and manage " (Harrot, Brace, and World in the central authorities that are used for following scheduling flow calculation task; New York; 1968) the sort of traditional project scheduling concept of discussing in the 569th page to the 584th page.

The development of process control system has following characteristics, and namely introducing and the management dullness at controlled system inner question rolls up, and this requirement is carried out corresponding relatively large logic to corresponding relatively large data in the time frame that can not prolong. The hardware that can carry out quickly logic and can store more data in the physical memory cell of an addressing can obtain above-mentioned feature to a certain extent, but takes this scheme to cost dearly. Solving system in future, effectively to expand them needed to this problem of comprehensive treatment capability of circumstances control be that a solution route really is provided, so that the computer of a large amount of distributions is carried out a large amount of logics with the interactive mode of time synchronized in whole network. The invention provides a kind of like this method and system.

The invention provides a kind of computer implemented method, can make the multicomputer node by interaction data communication with the cooperative mode actuating logic:

(a) alignment time defines so that each computer joint between computer node

Point accurately has identical " understanding " to the time;

(b) determine a communication scheduling table, this scheduling for each computer node

Table has been determined the different time for beginning in the future the corresponding communication of each time;

(c) to the corresponding dispatch list of each computer node transmission;

(d) according to the determined time of dispatch list begins each time communication separately.

The embodiment that describes provides a kind of computer network architectures, and it has a message carrier that links to each other with many computers; Clock system that is used for keeping the intercomputer time synchronized (can by all computer access in the network); With a scheduler that determines the computer to computer communication order, the time that it is carried out according to computer sets the tasks and causes that message carrier sets up communication at selected intercomputer and finish this function.

The embodiment that describes also provides a kind of time system of the overall situation read the reference value that is calibrated to a global definition time in this computer family.

The embodiment that describes also provides a kind of can make a plurality of member node with the computer implemented method of cooperative mode actuating logic.

The embodiment that describes also provides the simultaneity of intercomputer via message carrier communication.

The embodiment that describes also provides and has been the computer implemented method as a resource object scheduler task in a plurality of resource objects of collective's group.

The embodiment that describes also provides a kind of method, and different calculating functions is calibrated their timing definition by the frequency of oscillation of regulating their oscillators in this way.

Computer network architectures has the multi channel message carrier of support; A plurality of computers that are connected on the message carrier, every computer wherein all is programmed, and makes it cycle through a plurality of different conditions, comprises communications status, with definite task of time of implementation; One is used for keeping the time synchronized in the computer, can be by the clock system of all computer access; With a scheduler that is connected to the order of communicating by letter that is used for determining intercomputer on the message carrier, the time that it is carried out according to the computer task of determining causes that message carrier sets up communication at selected intercomputer and finish this function.

A computer node in a plurality of computer nodes (with the computing unit of central processing unit independently) is connected on the message carrier with other computer node, just can work in coordination with actuating logic with other computer node, also provide (a) master scheduling ability on a computer node to set up the master scheduling device; (b) in each computer node, determine this computer node required communication request table of communicating by letter with each time other computer node in the future time period of correspondence; (c) send communication request table via message carrier to the master scheduling device from each computer node; (d) each computer node via message carrier from master scheduling device received communication dispatch list; (e) each computer node is determined local schedule of tasks from the communication scheduling table that receives.

Said method makes a plurality of member node in the system carry out a large amount of logics with cooperative mode on the platform of a plurality of dispersions.

In one embodiment, the step of definition moment reference value is carried out by the master scheduling device in the above-mentioned implementation.

In one embodiment, in the above-mentioned implementation in individual member node the step of definition communication request table limited maximum communication quantity in the time interval in future, each time communication has a duration attribute, for obtaining valid function, this duration property value is less than a fixing maximum.

In one embodiment, the use of system of length of a game is so that all the different computer nodes (member node) in the system of expansion have the pervasive understanding to the time.

The method of scheduling comprises to the process as a resource (member) the object scheduling task in a plurality of resource objects of collective's group; This computer-implemented process (being called " apparent keystone resources (apparent critical resource, ACR) " process) comprises the following steps:

For this collective's group is determined a series of Future directions;

For each Future direction in the Future direction series is determined each resource object;

Determine corresponding to required reciprocation between any two resource objects of Future direction series;

Determine a usage example to finish each reciprocation for each resource object, this usage example has one and uses the duration property value;

Determine an access resources object and an accessed resource object corresponding to each usage example;

For each access resources object is determined total access time value, this value is the summation of all operating period property values of all usage examples when being the access resources object corresponding to this resource object;

For each accessed resource object is determined total accessed time value, this value is the summation of all operating period property values of all usage examples when being accessed resource object corresponding to this resource object;

Determine the continuous a series of future time sections in the dispatching cycle in dispatch list;

According to first the effective Future direction in this series Future direction, a Future reciprocation event scheduling that occurs between the following object is advanced in the earliest available time in first future time section:

(a) has the highest total access corresponding to this series Future direction

The value the access resources object and

(b) have the highest always interviewed corresponding to this series Future direction

The accessed resource object of the value of asking is (if at this access resources object and interviewed

Ask and have usage example-should future in dispatch list record between the resource object

Resource reciprocation event has with earliest available time in the dispatch list and equates

Zero-time, the access resources object corresponding with access resources object identification

Sign, the accessed resource object identification mark corresponding with accessed resource object

Will, and aforesaid operating period property value)

Deletion does not redefine Future direction series corresponding to the distinguishing mark of this access member object in the usage example of this accessed resource object in the Future reciprocation event of having dispatched and the Future reciprocation event of having dispatched so that those are also dispatched interactive Future direction from the scheduling candidate;

Repeat to determine total access value, determine total accessed value, scheduling future communications task and deletion be corresponding to the step of all usage examples of the resource object of having dispatched, until all can be scheduled all be scheduled into the usage example of first future time section till;

Carry out to determine circularly total access value, determine total accessed value, scheduling future communications task and deletion be corresponding to the step of all usage examples of predetermined resource object, and repeat these steps until all can be scheduled all is scheduled into the usage example of first future time section, and all resource reciprocation events all were scheduled into this dispatching cycle.

In one embodiment, message carrier has a plurality of passages that allow many communication events parallel, and the scheduling step is corresponding to a plurality of passages, and as a whole implementation, wherein earliest available time is corresponding to any one passage of this message carrier.

In certain embodiments, each member node has the variable-ratio oscillator, is used for adjusting the global clock calibration in local clock and the network. Also provide among these embodiment carried out by computer circuitry time calibration logic and oscillator frequency adjust logic to obtain the quick time calibration to special member node (computer node).

The message carrier of using in the above-described embodiments has (in a preferred embodiment) band crossbar switch or (in another embodiment) network topology with RF communication system.

Reading can be expressly understood other features and advantages of the invention to specific descriptions and the accompanying drawing of preferred embodiment more comprehensively.

Fig. 1 has represented the data flow diagram between the different units in the network that scatters.

Fig. 2 has described the computing unit network that scatters, and they are distributed in the whole world physically on the different position. This figure comprises the computing unit network, the master scheduling device and common time the source.

Fig. 3 has represented a bus-structured network diagram.

Fig. 4 has shown the network of annular layout in the computing environment of disperseing.

Fig. 5 has represented the network diagram of conversion star topology, and it is the preferred embodiment of network architecture.

Fig. 6 has represented the data structure schematic diagram of communicating by letter between computing unit and master scheduling device.

Fig. 7 has described the flow chart of master scheduling device to the implementation process of the communication scheduling between the computing unit of two separation.

Fig. 8 has represented the rough schematic of the scheduling logic (software that electronic circuit is performed) implemented in to the communication scheduling between the computing unit of two separation at the master scheduling device;

Fig. 9 has represented the detail flowchart of the scheduling logic that the master scheduling device is performed.

Figure 10 has represented to be described in be scheduled the first histogram of modulated degree communication between the rear computing unit of first connection.

Figure 11 has represented to describe the be scheduled histogram of the modulated degree communication between the rear computing unit of schedulable all connections in a first step.

Figure 12 has represented to be described in the be scheduled histogram of the modulated degree communication between the rear computing unit of all connections.

Figure 13 provides the detailed formation of message carrier, master scheduling device, clock and member node.

Figure 14 shown the technology that adopts in the preferred embodiment and those use in relation between the emphasis of most critical.

Figure 15 has represented that each computing unit is for realizing the flow chart of the logic that two inter-node communications are performed.

Figure 16 has described in a distribution network that realizes communication with radio frequency the schematic diagram of data flow between the variant unit.

As the introduction to this part explanation, system and method of the present invention is implemented: central authorities' scheduling of the task that cooperation is carried out, time is determined the operation of event, and the calibration of the upper corresponding oscillator of many computers (being also referred to as " member node ") on effective real-time synchronization and the network. In this, the network determined of this time comprise time-based rather than based on certain event (for example the reception of a signal or detect a specific bus can with) the communication between the computing unit that separates. Central authorities' scheduling of the task that cooperation is carried out, time is determined the operation of event, and effectively on real-time synchronization and the network on many computers the calibration of corresponding oscillator need computer to a real-time true mathematical logic presentation constantly share one accurate identical apparent. Suppose (this does not limit the scope of the invention) for instance and to this example, system-wide needs the precision of microsecond, this means, two different computers in this system must be known, when being 08:00:0000000 hour by the real-time moment of present networks authority's time source definition, the mathematical logic presentation that First in two computers has this real-time moment at 07:59:9999995 between the 08:00:0000005, and the mathematical logic presentation that in two computers second has this real-time moment also at 07:59:9999995 between the 08:00:0000005; As can concluding, two computers are not more than microsecond to the difference of the mathematical logic presentation in this real-time moment.

When the computer in the network of a distribution to a real-time true mathematical logic presentation constantly share one accurate identical when apparent, they just can be programmed according to the time and carry out specific task. The task of relating to many computers can realize central schedule, and for causing that with the method for determining the activity of time correlation lays a good foundation in the computer that scatters geographically. In this sense, the feature at a predetermined scheduled event of real-time moment generation is the certainty of real-time operation.

Use use that such time coordination also makes dynamic redundancy easily in the communication between computer. In dynamic redundancy, there are two or more computer to carry out separately identical logic, exchange the result of these operations, decide that alternatively these results that independently obtain think that further use obtains the associating conclusion of common understanding, and operation is with this result of joint effect. When the data that received by each computer were carried out the computer utility of same logic by other before result arbitration, just can carry out most effectively this process, so must there be like this some mechanism, lean on them jointly to determine the time that arbitration process begins. In the present invention, providing time synchronized between the redundant computer and scheduling is that (1) provides identical real-time constantly a mathematical logic presentation in computer about the mechanism of the task of a communication example; (2) when the communication example will occur, determine the moment of this computer with a kind of dispatching method in computer; And (3) then in involved each system of this communication example that occurs for the moment of communication exemplary definition, implement suitable transmission, reception and connection task be connected By this way, the moment that will carry out above mentioned arbitration task after having dispatched communication task, just pre-determined (with carried out communication task in reality after wait again the method for zero-time of arbitration task to be determined opposite).

This dispatching method is implemented by the operation dispatching logic. The concept of logic is relevant with software (computer program), and its purpose is to be reached in the process of carrying out this software by the electronic circuit system (usually being referred to as with term " computer ") that can carry out this software. The scheduling logic that is used to communication scheduling is based on the analysis that the order of the resource (for example time of implementation of the data space in the computer of appointment, each computer, each computer, the specified circuit in each computer) of having dispatched is done. This scheduling logic is used to the scheduling of network service, but it can also be applied to any one and need the scheduling of interactive resource between another, and it is particularly suitable for real-time application. The output that scheduling logic also is designed to be responsible for the scheduling of respective resources and can will produces from input (this is the set of connection request between any two computers the network) expectation it, i.e. dispatch list, the amount of needed time. For finishing this task, the scheduling logic scheduling has the available resources of maximum time requirements to generating dispatch list. The specific scheduling logic of describing in this explanation also is known as apparent keystone resources or ACR scheduling logic here, and following characteristic (1) estimation that it has reflected this operation is given the resource priority with peak demand in the resource that the each point of dispatch list has peak demand with (2). The special scheduling logic of describing is called as the ACR scheduling logic in this specification. In this, in a preferred embodiment, in case a specific task has been scheduled, it is not moved out of dispatch list to seek its replacer in dispatch list, and it is unduplicated that scheduling logic says in this sense. This quality is so that the preferential ACR scheduling logic type of scheduling logic is very suitable for real-time application just.

In the network communication dispatching of scheduling logic is used, considered the existence of transmission resources and reception resource. Each dispatch network communication, a transmission resources always links to each other with a reception resource.

In this network communication dispatching, master scheduling device and member node respectively have its duty.

Fig. 1 has represented the data flow diagram 130 of the associated data stream between general cell and the different general cell in the network that scatters, and shown a message carrier 100, first member node 107 and second member node 108, one is used for coordinating member node 107 via data channel 101 and data channel 102, the master scheduling device 106 of the communication between 108, be used for obtaining member node 107, synchronous clock selecting between 108 is (such as clock 109, with alternative clock 110) (via data channel 104, data channel 105, data channel 101, with data channel 102), and primary data table and the effectively required exchange of reciprocation in the computing environment of this distribution are (via via data channel 118, data channel 117, data channel 119, data channel 120, and data channel 121).

Clock 109 is distributed to member node (perhaps, message carrier 100 also can be distributed to member node 107,108 and master scheduling device 106 with alternative clock 110) by message carrier from the master scheduling device. To have a common time apparent with regard to allowing all to be connected in member node on the message carrier for this. In this, clock 109 has represented the earth around physics (as in the North America, in Europe, and in the Asia) clock of the overall situation used, alternative clock 110 then is represented as the clock of only being used by message carrier 100 of network design, on the other hand, the situation of serving as message carrier 100 by a computer is at random definition time of the computer that is attached thereto in the network that scatters. Clock 109 or clock 110 are periodically exported common clock signal via linking 104 or 105 accordingly.

Each member node 107,108 is responsible for receiving common clock signal; Effectively with this common clock signal " synchronously " (timing definition of node is calibrated in the very little tolerance with the timing definition of corresponding common clock signal, in order to according to the operation creativity system time synchronous with the clock frequency of having set up and fixing point); Receive suitable local event table 114,116 from master scheduling device 106; Formulate corresponding tasks carrying table according to the local event table 116,114 that receives and any attachment of a task that does not have to describe 114,116 kinds of local event tables; The execution list of executing the task; Produce communication request table 113,115; And send this tasks carrying table to master scheduling device 106. Each member node 107, the 108 as required task of node 107,108 produces connection request table 113, and 115, and by message carrier 100 these requests are passed to master scheduling device 106 in the aforementioned time. Master scheduling device 106 is the connection request table 113 that comes from member node, and 115 combine and put comprehensive required list 111 into, and it is used to produce the communication scheduling table of total event table 112 representatives. Then this total event table 112 is divided into the fragment that can be applied to each member node 107,108 and sends member node 107,108 to be stored into local event table 114,116 by message carrier 100 (also in the aforementioned time). Local event table 114,116 is used for suitable time correlation task by member node, comprise and the communicating by letter of other member node, time control.

Fig. 2 has described the computing unit network 220 that scatters, and they are distributed in the whole world physically on the different position, and it has shown example of the situation that data flow Figure 130 of Fig. 1 proposes. This figure comprises computer 201-214 and two networks that time source consists of, and these two time sources are that representative is corresponding to clock 203 and the clock 204 of the clock 109,110 of time flow graph; In this, Fig. 2 has shown the benchmark of clock 203 as all system times definition of North America and Eurasia, and clock 214 is then for determining a timing definition benchmark less than the computer 206,207,208,209 that is positioned at Africa of distribution on global and 214 main frame. Here the clock 109 in the DFD flow process of Fig. 1 of will having noted clock 203 " diagram ", and clock 214 " diagram " clock 110 in the DFD flow process of Fig. 1. In Fig. 2, described two kinds of the computing environment scattered may forms (helping the message carrier 100 of uniting, master scheduling device 106 and clock 104,105). The computing environment of scattering may be limited to a geographically limited zone, perhaps intersperses among the whole world. When interspersing among large geographic area, clock 203 may be a satellite time distribution system, such as global positioning system.

Fig. 3 has represented the network diagram of bus structures 330. Fig. 3 has shown the essential part among Fig. 1, and only message carrier 100 is the buses 300 that member node 107,108 is linked to each other with master scheduling device 106.

Fig. 4 has shown the computing environment of a dispersion, and wherein network is arranged to annular layout 430. Fig. 4 has shown the essential part among Fig. 1, and only message carrier 100 is by the arc 402,403,405 that effectively member node 107,108 and master scheduling device 106 is coupled together and 401 annulars that consist of.

Fig. 5 has represented the network diagram of conversion star topology 530, and it is preferential network architecture embodiment. Fig. 5 has also shown the essential part among Fig. 1, only message carrier 100 is a switching network that is arranged to star topology significantly, this layout couples together member node 107,108 and master scheduling device 106, and master scheduling device 106 is included in network switch 501 inside. This schematic diagram has shown preferential embodiment, because it can be by a large amount of I/O treating apparatus and the computing platform of dispersed placement are used to large-scale manufacture process (for finish the manufacture process that material is made at an equipment geographically, the movable part of this equipment uses the control signal from a computer control system to realize in real time change, and this system carries out the logic that reads metric data and output control signal from equipment). This star topology network 530 can pass through I/O (input signal/output signal) interface 510 for implementing manufacture process, I/O interface 522, I/O interface 519, I/O interface 512, I/O interface 513, I/0 interface 520,521 pairs of corresponding equipment of I/O interface read measurement data and output control signal. I/O interface 510, I/O interface 511, I/O interface 512 all links to each other with logger computer 505 with control computer 504 by LAN 515 with ring outlet 509 through ring outlet 508 with I/O interface 513. In a similar fashion, I/O interface 522, I/O interface 519, I/O interface 520 all links to each other with logger computer 507 with control computer 506 by LAN 514 with ring outlet 517 through ring outlet 518 with I/O interface 521. I/O interface 510, I/O interface 511, I/O interface 512, I/O interface 513, ring outlet 508, ring outlet 509, control computer 504, logger computer 505 and LAN 515 are with the cohesion operation of globality and the real time data logic highly of local logic. In a similar fashion, I/O interface 522, I/O interface 519, I/O interface 520, and I/O interface 521, ring outlet 518, ring outlet 517, control computer 506, logger computer 507 and LAN 514 are with the cohesion operation of globality and the real time data logic highly of local logic. Network switch 501 so that

(a) the I/O interface 510, I/O interface 511, I/O interface 512, I/O

Interface 513, ring outlet 508, ring outlet 509, control computer

504, logger computer 505 and LAN 515

With

(b) the I/O interface 522, I/O interface 519, and I/O interface 520, and

I/O interface 521, ring outlet 518, ring outlet 517, control computer

506, the cohesion between logger computer 507 and the LAN 514 can send, and this network switch is implemented (1) to the communication of the on all four mathematical logic presentation in a moment in providing in real time in the computer; (2) at computer 506, logger computer 507, control computer 504, logger computer 505, in the user interface 503 in the time of any two each time communication example generations through using network switch 501 between each, be control computer 506, logger computer 507, control computer 504, logger computer 505, user interface 503, network support 502, and network switch 501 is determined the program of each master scheduling device 106 constantly; (3) in the time of determining for the communication example the related parts of this communication example are in time sent, receive and the link task. As single information source, acceptable communication parter in network support 502 definite star topologies 530 also adds up and the diagnostic signal in star topology 530 provides access, and it is collaborative with conversion star topology 530 that it is used as a node. Network switch 501 grid of reference supports 502 are listed acceptable communication parter in the star topology 530 and are confirmed that those are determining to send and receive the example needs of communicating by letter that run counter to this tabulation when linking task.

In one embodiment, I/O interface 522, I/O interface 519, I/O interface 520, and I/O interface 521 be easy to by one such as United States Patent (USP) the 5th, 555, the described process control system of 424 (24Sederlund etc.) is realized. This on September 10th, 1996 issued patents authorize Sederlund, Edward R.; Lindesmith, Robert J.; Root, Larry A.; Dupree, Wayne P.; And Thomas, " the Harvard architecture computer memory system with expansion of addressable increment " of Lowell V.. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

In an embodiment more preferably, I/O interface 522, I/O interface 519, I/O interface 520, be easy to via a redundant process control computer system in the process control system two processes control of use computers (such as described process control computers such as 24Sederlund) with I/O interface 521, such as United States Patent (USP) the 5th, 583,757 (Baca, Jr. etc.) described system realizes, this on December 10th, 1996 issued patents authorize Baca, Jr.Eloy; Dupree, Wayne P.; Grinwis, Donald J.; Kanse, Johannes C.; Pelletier, Douglas P.; And Schlze, " the input signal resolution method that is used for dynamic redundancy process control computer " of Oscar E.. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

Authorized Sederlund, Edward R. in the United States Patent (USP) 5,568,615 (15Sederlund etc.) that on October 22nd, 1996 authorized; Thomas, Nadene T.; Lindesmith, Robert J.; And Cowles, " the Stealth interface that is used for process control computer " of Russel W. has wherein been described an embodiment who obtains data access for described process control computers such as 24Sederlund. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

Authorize Glaser, Robert S. in the United States Patent (USP) 5,428,769 (69Glaser etc.) that authorize June 27 nineteen ninety-five; Hoy, Robert S.; Fernandez, G.Paul; And Grai, and Timothy J. " the process control interface system that unit, triplex level redundancy far field is arranged " wherein described an embodiment of a system that the unit, far field is provided that uses with described process control computers such as 24Sederlund. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

The United States Patent (USP) 5 of authorizing on September 29th, 1992,151,866 (66Glaser etc.) authorize Glaser, R.Steven and Bade, an embodiment of a system that the interface of electric current in the readout power distribution system is provided that uses with unit, described far field such as 69Glaser has wherein been described in " the high-speed power analysis " of Jeffrey M.. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

In following United States Patent (USP)

(a) in the Allbery that authorizes of mandate on May 21st, 1996, Jr.,

James D.；Troisi,Peter.,A.；Johnson,Susan J.；Cullen,

James H.；Butler,Richard L.；Ferreira,James P.；Ellison,

Joseph; Patel, Chiman L.; Uban, James E.; And Schultz, Dale

H. " the intelligent process control of the time calibration that can carry out corresponding data series

Communication system and method " United States Patent (USP) 5,519,603 (Allbery, Jr.

Deng);

(b) in June 27 nineteen ninety-five authorize authorize Bruijn, Ronny

P.；Verboven,Marc L.K.；van Weele,Leonardus A.；

Vermeire, Roger R.; Schultz, Oscar E.; Schultz, Dale H.; With

Bell, Brian G. " use new random number be used for tie in the call duration time time limit

Rebulid between First and second computer peace of communication in limited time before the bundle

The full communication system " United States Patent (USP) 5,428,745 (45de Bruijn,

Deng); With

(c) on October 1st, 1996 authorize authorize de Bruijn,

Ronney P.；van Weele,Leonardus A.；Verboven,Marc L.K.；

Vermeire, Roger R.; Schultz, Oscar E.; Bell, Brian G. and

Schultz, Dale H. " by check with order in variable identifier

The state of permission indicator determine whether giving an order controlling computer

System and method for " United States Patent (USP) 5,561,770 (70de Bruijn,

Deng); The embodiment of a system that high speed routing is provided that uses with redundant process control computer system has been described.

Here specially these patent citations are incorporated in this application and obtain and use method of the present invention to show.

Authorize van Weele, Leonardus A. in the United States Patent (USP) 5,631,825 (van Weele etc.) that on May 20th, 1997 authorized; De Bruijn, Ronney P.; Vermeire, Roger R.; Zemiering, Christo; And Lenting, " for the manufacture of the active station of process control system " of Ben wherein described one with the unify embodiment of system of the interface that the manward is provided that the interface described in the above-mentioned patent uses of redundant process control department of computer science. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show. Although this manward's interface can be by I/O interface 522, I/O interface 519, I/O interface 520, with in the I/O interface 521 each or any one use as user interface, still should be noted that in the special case of conversion star topology 530 user interface 503 is what to be preferably through network switch 501 to the Interface design of whole system.

(a) on February 13rd, 1996 authorize authorize Pressnall, Dana W.; Polishak, Jeffery T.; Felix, Bradley K.; Durisin, Michael J.; And Ellison, the United States Patent (USP) 5 of Joseph " information display system that is used for the control of dynamic redundancy computerized processes ", 491,625 (Pressnall etc.) and (b) in April 18 nineteen ninety-five authorize authorize Van de Lavior, Ronny; Follon, Marinus (Neerpelt, BE); And Ravenscroft, the United States Patent (USP) 5 of Ian " information display system that is used for the control of dynamic redundancy computerized processes ", among 408,603 (the Van de Lavior etc.) embodiment who the application code of carrying out is provided the system of real-time interpretation in said system has been described. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

One provides the large-scale image overview system in the face of human interface to be applied in this preferred embodiment with active station. In the United States Patent (USP) of authorizing John L. Clement 5,726,668 (Clement) of authorizing on March 10th, 1998, this large-scale image overview system has been described. Here specially this patent citation is incorporated in this application and obtains and use method of the present invention to show.

In the prepreerence embodiment of generalities, with one by Wayne Dupree, Jeffery Lucas, Larry Root, Gerrit Verniers, and Stephen Churchill " computer of a private processes circulation (Context-Cycling) " by name that file on February 12nd, 1997 08/797, the private processes cycle calculations machine of describing in No. 967 U.S. Patent applications, I/O interface 510, I/O interface 511, I/O interface 522, I/O interface 519, I/O interface 512, I/O interface 513, I/O interface 520, I/O interface 521, ring outlet 508, ring outlet 509, control computer 504, logger computer 505, ring outlet 518, ring outlet 517, control computer 506, logger computer 507, I/O interface 522, I/O interface 519, I/O interface 520, I/O interface 52l, ring outlet 518, ring outlet 517, control computer 506, network support 502, network switch 501, logger computer 507 is used in succession. This private processes cycle calculations machine is with the system that 69Glaser with the interface of field detector etc. is provided. Here specially this patent is used to quote to be incorporated in this application and obtained and use method of the present invention to show.

This flow process circulation treatment device provides specified loop to a large amount of with different input and output loops, and each specified loop has a specific record device at least with process and the configuration status of relevant relevant resource, attribute or outward appearance of maintenance when another flow process is in current state and loop or function; The combination that specified loop also is provided and sets up the specific record device of Circuits System according to a flow process. Flow process also comprises specific interrelated logic for Circuits System. A plurality of flow processs are shared public resource in CPU.

Authorize Dupree, Wayne P. in the United States Patent (USP) 5,655,133 (Dupree etc.) that on August 5th, 1997 authorized; Churchill, Stephen G.; Gallant Jeffery R.; Root, Larry A.; Bressette, William J.; Orr, III, Robert A.; Ramaswamy, Srikala; Lucas, Jeffery A.; And Bleck, " computer of extensive SuperScale Harvard architecture " of James wherein described the coprocessor with the common needs that use of organization procedure cycle calculations machine. Here specially this patent citation is incorporated in this application to show the method that obtains and use preferential expectation of the present invention. Network switch 501 also preferential the use can be implemented at least 99 crossbar switch (Xbar) 516 parallel or continuous communiction links.

In network switch 501, provide common clock signal (being the peek of clock 109 or the demonstration of clock 110) with master scheduling device 106, and responsible clock signal is distributed; From each member node 107,108 (in this special case of network 530, control computer 506, logger computer 507, control computer 504, logger computer 505, user interface 503, network support 502 and network switch 501 all are the member node in the flow of data stream 107 and 108 examples among Fig. 1) collect request; Be each member node 107,108 dispatching communication task; Send dispatch list to each member node 107,108; And the connection between the enforcement member node 107,108.

The purpose of distributing a common clock signal is to make each node 108,107 on the network 530 have identical understanding to the time. The identical understanding that has the time is vital in the present invention, carries out because the scheduling of task is based on the time. Therefore, the common clock signal is used for effectively all nodes 108,107 being synchronized to the time of sharing. On the basis of predetermined period, the master scheduling device 106 in the network switch 501 sends a message to each beginning with definite time cycle in the node 108,107.

Each dispatching cycle is from each member node 107,108 (control computer 506, logger computer 507, control computer 504, logger computer 505, user interface 503, any one in network support 502 and the network switch 501) connection request of collection. Master scheduling device 106 in each network switch 501 all sets apart to receive from each member node 107 dispatching cycle, 108 (control computers 506, logger computer 507, control computer 504, logger computer 505, user interface 503, any one in network support 502 and the network switch 501) connection request. Communication request comes from the member node 107,108 of the requested communication of needs. Master scheduling device 106 in the network switch 501 receives each request and consider these requests in its scheduling logic. Each request comprises the information of source, target and the length aspect of communication. In addition, each request also may comprise priority indicator and send the required information of its information to the application-specific of carrying out at destination node. After the task of collecting connection request from all member node 107,108 was finished, its result was collected into a comprehensive required list 111. Attention comprehensive required list 111 entrances may be in network switch 501 master scheduling device 106 internal initialization itself; Specific connection request also may have repeat property, is for example caused repeatedly by single signal to connect in a series of time cycles in future.

Then the following communication that will occur of 106 scheduling of the master scheduling device in the network switch 501. Although the apparent keystone resources of discussing is preferential method in the present note, it is not conclusive dispatching employed method for executive communication in determining the function of network in the time. Logic is given the allocation of communications time that needs. Can determine in dispatching cycle, when to stop scheduling process with following three kinds of modes: the first, all communication request that can be scheduled all are scheduled, and therefore do not have remaining scheduler task; The second, there is not time (processor must stop scheduling process to continue its next task) of remaining operation dispatching; The 3rd, existence is more than the required connection (bandwidth exhaustion) that can dispatch in available scheduling window, and superfluous request is maintained at and waits for scheduling trial next time in the formation in this case.

The scheduling that occurs is deposited into total event table 112. This total event table 112 except comprise with comprehensive required list 111 in the same information, also comprise the zero-time 621 of the communication of having dispatched. Note, may load in advance certain task to walk around scheduling process, as dispatching requirement between the special time that strengthens task is determined is needed in total event table 112.

Then the master scheduling device 106 in the network switch 501 transmits local event table 114,116 (part of dispatch list, it is suitable for specific member node 107,108) to each corresponding member node. Appear at and all have entrance at local event table 114,116 with member node 107,108 as each event of source or target in total event table 112. Any one relates to the member node 107 of any communication, 108 (control computers 506, logger computer 507, control computer 504, logger computer 505, user interface 503, any one in network support 502 and the network switch 501) receive its local event table 114,116.

Then the master scheduling device in the network switch 501 106 was implemented to connect according to the time that is assigned with that appears in total event table 112. When using the preferred embodiment of clog-free conversion, the process of these connections causes occurring simultaneously effective situation of a plurality of connections.

The comprehensive required list 111 that is arranged in the master scheduling device 106 of network switch 501 may comprise does not have Iterim Change ground to implement the continuous connection that repeats, the connection that perhaps dynamically needs (just implement when needed, usually need to be proposed by a node on the network). This connection request table comprises a source designator, target designator and a duration designator (seeing the Fig. 6 that shows record details 600) at least. The parameter relevant with scheduling in addition optionally comprises the designator as priority indicator 617.

Fig. 6 has shown and has been used for dispatching and the original table of task coordinate process and the combination 600 of information. Connection request table 115 member node 107,108 neutralization in the master scheduling device 106 in the network switch 501 when being combined into comprehensive required list 111, all have source 614, target 615, the zone 610 of length 616 and possible priority 617. The source 618 of data in these zones 610 of scheduler logic 611 usefulness in master scheduling device 106 to come with the form of zone total event table 112 of 612 to connect as individual time, target 619, length 620 and zero-time 621 structure dispatch lists. Note scheduler logic 611 extraction source 614 from connection request table 111, target 615 and length 616 are although zero-time 621 is products of scheduling process itself.

Local event table 114,116 is the simple subsystem of total event table 112, total event table 112 is divided into those and can be applicable to corresponding member node 107,108 (control computers 506, logger computer 507, control computer 504, logger computer 505, user interface 503, any one in network support 502 and the network switch 501) part-do not need any operation is done in zone 613. Tunnel 601,602,603,606,607 and 608 have shown the information flow from connection request table section 610 to local event table zone 613. Tunnel 604 has shown the input of the priority 617 of dispatch list; Tunnel 609 has shown from the next zero-time 621,625 of scheduler logic 611 transmission.

Then these data of scheduler logic 611 usefulness draw dispatch list so that the connection that needs to be provided. The result of the scheduling process of scheduler logic 611 is recorded in total event table 112, and total event table again then be carried out syntax and divide and enter in the corresponding local event table 116 to distribute to network node (seeing Fig. 1). These event tables comprise source 622, target 623, the zone 613 of duration 624 and zero-time 625.

In network node, local event table 114,116 combine with local task list and actuating logic to guarantee to finish all work that will do and all time schedulings corresponding to (or with it without conflicting) device logic 611 appointments that are scheduled.

Fig. 7 has summarized general scheduling process 710. Collect connection request at predetermined request receiving in cycle. Through week predetermined time after date, must process a large amount of connection requests. More connection is not under an embargo, but must set up a point of cut-off. Fig. 7 has shown the needed basic step of scheduling connection request in the master scheduling device 106 in network switch 501. First three step 701-703 provides the timing definition of calibration for all nodes. In step 704, receive connection request from all member node 107,108 (these nodes send their once communication of request conduct from preferential operation in the predetermined moment of their local schedule of tasks). In step 705, when scheduling process was applied to those requests that those collect from member node 107,108 in special time period windows that expression finishes, the connection request process stopped so that can be for determining dispatch list future. In case request is collected (seeing Fig. 8 for details), the step 706 of scheduling connection request just can begin. In the step 707, in case finished this scheduling, just from total event table 112, extract suitable part and create local event table 114,116, then in step 708, pass to node 107,108 in order in step 709, implementing.

Generally for the computing environment of scattering, may need at any time between the node 107,108 to connect, this means that connection request may start from any moment, and must allow some to process the service time of these connections. There is Four processes, and in network 530, repeat with the speed consistent with the response of wanting: collect connection request, the scheduling connection request distributes the dispatch list that produces to network node, by the required connection communication (see Table 1) of network 530 through having set up.

Four scheduling processes of table 1-

Request connects
	Scheduling connects
Transmit scheduling
	Communication

Table 2 has shown the time relationship of this Four processes, and as statement, this Four processes is carrying out always.

Table 2-scheduling flow

Request connects
				Request connects	Scheduling connects
Request connects	Scheduling connects	Transmit scheduling A
				Request connects	Scheduling connects	Transmit scheduling B	Communication A
Request connects	Scheduling connects	Transmit scheduling C	Communication B
				Request connects	Scheduling connects	Transmit scheduling D	Communication C
Request connects	Scheduling connects	Transmit scheduling E	Communication D
				Request connects	Scheduling connects	Transmit scheduling F	Communication E

Attention is communicated by letter in the row of table 2 rightmost side and is always occurred in as after the repeating several times at first of connection setup basis. To any one the specific communication in a time period window, all in the time period of its front window, transmit effective communication scheduling table, and be scheduled in the time period window of request before the dispatch list transmission, and these requests before scheduling to be proposed. Can life period in this flow process variation in the scheduling, but individual flow process this is similar to the flow process in the preferred embodiment.

The design of the logic of using in the currently preferred embodiment is with two purposes: must obtain the efficient scheduling table, and produce the required period of this dispatch list and must be limited in rational level. Effectively scheduling relates to one based on the simple procedure of following prerequisite, namely the efficient of integrated dispatch table at first be an efficient function so that in this dispatch list resource utilized as much as possible. This means and to guarantee always that in scheduling process the resource with the highest pending requirement obtains priority scheduling. For obtain dispatch list in the acceptable time, logic is designed to each request and is scheduled once and only is scheduled once (trial and error-logic are considered to unacceptable), and to the limited processing requirements of each connection request.

As shown in Figure 8, the basic process in the scheduling 806 is as follows:

1) definite criterion of weighing resource requirement;

2) according to key with resource classification;

3) provide the company that can provide in this step by the critical order of resource

Connect;

4) further finish processed task thereby vacate available of front

Resource;

5) repeating step 2-4 is until all requests are all processed or until do not have machine

Can process till more requests.

In this scheduling process, any request that can not be processed can be suggested in next scheduling process again.

The flow chart of Fig. 8 has been summarized the process of using in the scheduling logic of an apparent keystone resources. The first step is to determine to weigh the criterion 800 of resource use amount, and for this preferred embodiment, the sending and receiving passage is exactly the resource of having dispatched, and the use duration of each resource just is chosen as criterion.

In case determined the use amount of resource, just classified by use amount 801. This allows to use maximum resources at first to be dispatched in step 802. When all resources that can be scheduled all have been scheduled (connection of existing request may not need or use when not allowing all resources), the time period window reach of scheduling in step 803 until all available resources all available till. If the time that also has more need connection request to be processed and in dispatch list, process in addition these requests, determined corresponding to after this problem of repetition of step 801 to 803, just returning this process of repetition to whether needing once in step 804. According to the decision of the completed step 804 of scheduling example, the subprocess 806 of scheduler logic proceeds to step 805 and withdraws from.

Fig. 9 has represented the more detailed general picture of subprocess 950 of scheduling logic of the preferred embodiment of apparent keystone resources (ACR). As the preferential enforcement of process shown in Figure 8, this sketch centers on the step of mentioning among this Fig. 8 and has increased some conditional operations and the required action of Exception handling. The starting point of the ACR process that step 901 representative is preferential. In step 902, set up the ACR table, in step 903, according to the definite not existence of dispatch request of the SAV value in the ACR table (if should value indicate not request in the ACR table, just execution in step 918 withdraws from immediately). If there is request, ACR subprocess 950 carries out step 904 calculation window size, and step 905 is searched the request of appointment, and step 906 checks the request of appointment and the existence of communicating pair. If should ask and the communicating pair existence, this process is just carried out step 912 and is added this event in the event table and adjustment ACR path in step 913 before the next request in carry out step 914. If the high value in the step 915 is zero, the definite of (future) time just moves forward in step 916 among the ACR, step 917 checks that bandwidth has time enough further to dispatch with affirmation, then returns step 914 for next request process in these management and running. If in step 906, the recipient of communication does not exist, and just definition substitutes the recipient and estimates in step 908 in step 907. If step 908 less than the alternative recipient of definition, is just selected an alternative sender and is estimated in step 920 in step 919; If can accept (step 920), recipient and process that step 921 is selected to have the highest ACR value return 908; Unacceptable if (step 920), process enters step 916. When having defined the replacer in the step 908, process is via step 909, and 910 and 911 dispatch to 912. Consider that of showing corresponding to table 3 comprises the example of communication request of the hypothesis network of five nodes, can describe the virtual data logical process of carrying out by above-mentioned steps well by this example.

Table 3: initial request table

The source	Target	Length	Priority
					1	2	15	1
1	4	23	1
				1	5	2	1
2	1	24	1
				3	1	16	1
3	1	12	1
				3	2	20	1
3	3	8	1
				3	4	19	1
3	4	16	1
				3	5	4	1
3	5	9	1
				4	1	13	1
4	5	9	1
				5	1	21	1
5	1	8	1
				5	2	8	1
5	3	1	1
				5	4	18	1
				Amount to		246

Suppose that length measures with random time unit, may be millisecond, and n different preferential level arranged that from 1 to n, wherein 1 is the most urgent and n is last.

At first, attention broadcast-type net (wherein all interbehaviors all need to share bandwidth, for example have carrier sense multiple access or the CSMA/CD network of the detection of hitting) is finished above-mentioned request required service time of minimum in theory needs 246 chronomeres.

The introducing of nonblocking switching device allows five transmission resources and five reception resources to move simultaneously. Be the benefit of analytical applications nonblocking switching device, the requirement on each resource of one of this routine scheduler logic model table assessment, such as table 4:

Table 4: initially gather access value

X/R		1	2	3	4	5		SAV
1		-	15	-	23	2		40
									2		24	-	-	-	-		24
3		28	20	8	35	13		104
									4		13	-	-	-	9		22
5		29	8	1	18	-		56
SAV		94	43	9	76	24		246

The access value that amounts to, or SAV just are the summation of row or column of sender's (row of X-table) and recipient's's (row that R-shows) reciprocation number of times simply, and have reflected the summation of the requirement of measuring with random time unit on special resource. Note selecting the required theoretical minimum time of nonblocking switching device post processing request to drop to 104 chronomeres from 246 chronomeres. Because sender 3 has the highest SAV, and when sender 3 was processed, other sender usually also can be processed, so the minimum service time of these requests is restricted by sender 3. When theory lower bound may be for 104 chronomere, the actual minimum service time may be grown and even may depend on and connect processed order. But next target is to determine the optimal scheduling method so that the ability maximization (minimizing required service time) of existing network.

To this special example hypothesis, synthetic dispatch list is applied to the time period window of future time, no matter composition and the quantity of request, this window all has 150 chronomere's durations. This example starts from checking in time=0 connection of o'clock carrying out for the dispatch list of this 150 unit. First aim is judged by SAV, is those requests relevant with sender 3 and recipient 1 of scheduling. With reference to the initial request table of front, this routine scheduler logic finds the relevant request of that have 16 chronomere's durations and a sender 3 and recipient 1. Because this routine scheduler logic is the connection in scheduling time=0 o'clock just, and this routine scheduler logic has 150 unit durations, and this routine scheduler logic is put 16 word requests into the time period scheduling window, and reciprocation is recorded into event table table 5.

Table 5: intermediate event table

The source	Target	Length	Priority	Zero-time
					3	1	16	1	0
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-

Table 5 can preferentially be expressed as the Figure 100 1 among Figure 10. The task of having dispatched is expressed in sender 3 unit 1002 of having dispatched and recipient 3 unit 1003 of having dispatched.

Then required list necessarily reflects the fact that last request is processed in some way. This adjusted required list may be similar to table 6.

Table 6: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	-
1	4	23	1	-
					1	5	2	1	-
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	-
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-
3	4	16	1	-
					3	5	4	1	-
3	5	9	1	-
					4	1	13	1	-
4	5	9	1	-
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	-
5	3	1	1	-
					5	4	18	1	-
					Amount to		246

Because sender 3 and recipient 1 are now o'clock occupied in time=0, their state must be pointed out this point, so the SAV that upgrades table has some changes thereupon, may be similar to table 7 now.

Table 7: access value is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	40	78
			2	24	43
3	88	9
			4	22	76
5	56	24

The SAV that notes sender 3 and recipient 1 has been adjusted to reflect the request of having processed. This is because scheduler logic continues to give to hit in the dispatch list the highest scheduling of resource priority. This process be other resource need to be for cost, and the poor efficiency that is placed in some cases the resource with low SAV can cause these resources to have than hit the highest taller the hitting of those resources when initial at the remainder of dispatch list. For this reason, this agreement is called as apparent keystone resources scheduling (dispatching priority always appears at resource on the critical path to those in the moment of considering), or the ACR scheduling.

Significantly, this routine scheduler logic should be attempted to dispatch from sender 5 to the recipient 4 request, both following two remaining schedulable relevant with the scheduling of hitting simultaneous resource. Since there is such request, so the execution of this routine scheduler logic is similar to table 8 so that those show, 9 and 10.

Table 8: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	-
1	4	23	1	-
					1	5	2	1	-
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	-
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-
3	4	16	1	-
					3	5	4	1	-
3	5	9	1	-
					4	1	13	1	-
4	5	9	1	-
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	-
5	3	1	1	-
					5	4	18	1	0
					Amount to		246

Table 9: intermediate event table

The source	Target	Length	Priority	Zero-time
					3	1	16	1	0
5	4	18	1	0
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-

Table 10: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	40	78
			2	24	43
3	88	9
			4	22	58
5	38	24

Notice that this event table obviously is (in chronological order) classification chart from the next request of having dispatched of required list. For this reason, each interstage of dispatching step is not revealed.

Repeat the process of front, the communication example request (i.e. 2 the connection from sender 1 to the recipient) that this routine scheduler logic scheduling is 1 to 2, and these tables are adjusted to following table 11 and 12:

Table 11: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	0
1	4	23	1	-
					1	5	2	1	-
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	-
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-

3	4	16	1	-
					3	5	4	1	-
3	5	9	1	-
					4	1	13	1	-
4	5	9	1	-
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	-
5	3	1	1	-
					5	4	18	1	0
					Amount to		246

Table 12: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	25	78
			2	24	28
3	88	9
			4	22	58
5	38	24

Next this routine scheduler logic searches 2 to 5 communication example request, but does not make such request. Below determining, had several selections by in the combination of attempting, but preferred plan is to select according to the balance between bandwidth efficiency and the disposal cost. (having minimum bandwidth hits, a plurality of simple rules can be selected, to this preferred embodiment, this special sender be left in the basket and next sender selected after, for existing sender selects to substitute the recipient until all recipients' selection is depleted, and hit the highest recipient be reconsidered till-see Fig. 9. )

Use this logic, search 2 to 3 communication example request but do not find, then 4 to 5 communication example request searched, found and selected. SAV table is updated, and determines in time=0 connection that o'clock do not have other to carry out. Event signature is being shown as table 13.

Table 13: intermediate event table

The source	Target	Length	Priority	Zero-time
					3	1	16	1	0
5	4	18	1	0
					1	2	15	1	0
4	5	9	1	0
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-
-	-	-	-	-
					-	-	-	-	-

These four connections also can be represented visually by the scheduling graph 1101 among Figure 11. The sender unit 1002 of having dispatched among Figure 10 and the recipient unit 1003 of having dispatched correspondingly are expressed as the unit 1102 and 1103 of having dispatched, and added the unit 1104 dispatched and be connected (4 to 5 connections), 1106 with are connected (1 to 2 connection) and 1108 with are connected (4 to 5 connections).

The logic that relates to front shift time equals to be advanced in chronological order next end. Here 4 to 5 communication example request is exactly chronological next end, thus before this example shift time then between=9 o'clock, constantly sender 4 and recipient 5 just can use again at this. Required list and SAV table is displayed in table 14 and 15 now.

Table 14: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	0
1	4	23	1	-
					1	5	2	1	-
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	-
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-
3	4	16	1	-
					3	5	4	1	-
3	5	9	1	0
					4	1	13	1	-
4	5	9	1	-
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	-
5	3	1	1	-
					5	4	18	1	0
					Amount to		246

Table 15: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	25	78
			2	24	28
3	88	9
			4	13	58
5	38	15

In the communication example request of searching 2 to 5,2 to 3 communication example request, 4 to 5 communication example, after request 4 to 3 communication example request and 4 to 3 the example request of communicating by letter, this routine scheduler logic judge this example must be again before shift time-in time=9, have virtually no request can be scheduled. Finish now is in time=15 o'clock next time, this constantly 1 to 2 communication example connect and finish and the SAV table is table 16.

Table 16: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	25	78
			2	24	28
3	88	9
			4	13	58
5	38	15

This routine scheduler logic is searched now 1 to 2 communication example request, but 1 to 2 communication example has been scheduled, so the communication example request of this example search 1 to 5, this request has been found and dispatched to this example. New SAV table has been exactly table 17.

Table 17: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	23	78
			2	24	28
3	88	9
			4	13	58
5	38	13

Then this routine scheduler logic searches 2 to 2 communication example request, 2 to 3 communication example request, 4 to 2 communication example request and 4 to 3 the example request of communicating by letter before the shift time before determining again. Shift time from 15 to 16 before this example seems such as table 18 at this moment SAV table.

Table 18: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	23	78
			2	24	28
3	88	9
			4	13	58
5	38	13

Then this routine scheduler logic dispatches 3 to 1 communication example request, and it is proved to be o'clock unique request that can be scheduled in time=16. Shift time is to time=17 before this routine scheduler logic behaviour, and these tables seem such as table 19 20 and 21.

Table 19: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	23	66
			2	24	28
3	76	9
			4	13	58
5	38	13

Table 20: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	0
1	4	23	1	-
					1	5	2	1	15
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	16
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-
3	4	16	1	-
					3	5	4	1	-
3	5	9	1	-
					4	1	13	1	-
4	5	9	1	0
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	-
5	3	1	1	-
					5	4	18	1	0
					Amount to		246

O'clock do not have the request that can be scheduled in time=17, so shift time is to time=18 before this example, 5 to 4 communication example request finishes this moment.

Table 21: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	23	66
			2	24	28
3	76	9
			4	13	58
5	38	13

After unsuccessful to the search of 5 to 4 communication example request, found 5 to 2 communication example request and 1 to 4 the example request of communicating by letter, and these tables seem such as table 22 and 23:

Table 22: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	66
			2	24	20
3	76	9
			4	13	35
5	30	13

Table 23: intermediate request table

The source	Target	Length	Priority	Zero-time
					1	2	15	1	0
1	4	23	1	18
					1	5	2	1	15
2	1	24	1	-
					3	1	16	1	0
3	1	12	1	16
					3	2	20	1	-
3	3	8	1	-
					3	4	19	1	-
3	4	16	1	-
					3	5	4	1	-
3	5	9	1	-
					4	1	13	1	-
4	5	9	1	0
					5	1	21	1	-
5	1	8	1	-
					5	2	8	1	18
5	3	1	1	-
					5	4	18	1	0
					Amount to		246

Shift time is to time=26 before this routine scheduler logic behaviour, and this moment, this routine scheduler logic was grasped definition list 24:

Table 24: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	66
			2	24	20
3	76	9
			4	13	35
5	30	13

This routine scheduler logic is the communication example request of scheduling 5 to 3 now, and it is all requests that constantly can be scheduled at this. Then the time is moved forward to time=27, and then 28, and the SAV table seems such as table 25.

Table 25: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	66
			2	24	20
3	76	8
			4	13	35
5	29	13

Since there is not 3 to 1 communication example request, the communication example request of the scheduler logic scheduling 3 to 2 that this is routine and 5 to 1 the example request of communicating by letter, and front shift time to 31, and then to 48, the SAV table becomes table 26 this moment.

Table 26: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	45
			2	24	-
3	37	8
			4	13	16
5	8	13

The communication example request of the scheduler logic scheduling 3 to 4 that this is routine, and front shift time to 49, the result provides table 27.

Table 27: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	45
			2	24	-
3	37	8
			4	13	16
5	8	13

The communication example request of this routine scheduler logic predetermined 2 to 1, and front shift time to 67 provide table 28.

Table 28: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	21
			2	-
3	29	8
			4	13	16
5	8	13

The communication example request of now scheduling 3 to 3 of this routine scheduler logic and 5 to 1 the example request of communicating by letter, and front shift time to 73, shown in table 29.

Table 29: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	21
			2	-	-
3	29	-
			4	13	16
5	8	13

Then this routine scheduler logic dispatches 4 to 1 communication example request, and front shift time to 75, causes table 30.

Table 30: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	8
			2	-	-
3	29	-
			4	-	16
5	8	13

This allows this routine scheduler logic to dispatch remaining 3 to 4 communication example request, and shift time causes the data of table 31 to time=86 before this routine scheduler logic after this.

Table 31: the access value table is gathered in the centre

Passage	Sender SAV	Recipient SAV
1	-	8
			2	-	-
3	13	-
			4	-	-
5	8	13

This routine scheduler logic is the communication example request of scheduling 5 to 1 now, and front shift time to 91 is dispatched first communication example request of 3 to 5, and front shift time to 94 then to 95, and is dispatched last request, another communication example request of 3 to 5. The result event table seems such as table 32.

Table 32: final event table

The source	Target	Length	Priority	Zero-time
					3	1	16	1	0
5	4	18	1	0
					1	2	15	1	0
4	5	9	1	0
					1	5	2	1	15
3	1	12	1	16
					5	2	8	1	18
1	4	23	1	18
					5	3	1	1	26
3	2	20	1	28
					5	1	21	1	28
3	4	19	1	48
					2	1	24	1	49
3	3	8	1	67
					4	1	13	1	73
3	4	16	1	75
					5	1	8	1	86
3	5	4	1	91
					3	5	9	1	95

Note whole required list after 104 chronomeres, finish-this is to process in theory required minimum time amount of these specific requests.

Final dispatch list is shown as the scheduling graph 1201 among Figure 12 with graphics mode. The sender unit 1002 of having dispatched among Figure 10 and the recipient unit 1003 of having dispatched correspondingly are expressed as the unit 1202 and 1203 of having dispatched; The unit of having dispatched 1104 among Figure 11 and 1105,1106 and 1107 and 1108 and 1109 correspondingly is expressed as the unit 1204,1205,1206,1207,1208 and 1209 of having dispatched; And shown the 1216-1217 of the communication example of new unit correspondence 1 to 4, the 1214-1215 of corresponding 5 to 2 communication example, the 1212-1213 of corresponding 3 to 1 communication example, the 1220-1221 of corresponding 3 to 2 communication example, the 1222-1223 of corresponding 5 to 1 communication example, the 1226-1227 of corresponding 2 to 1 communication example, the 1224-1225 of corresponding 3 to 4 communication example, the 1228-1229 of corresponding 3 to 3 communication example, the 1230-1231 of corresponding 4 to 1 communication example, the 1232-1233 of corresponding 3 to 4 communication example, the 1234-1235 of corresponding 5 to 1 communication example is with the 1238-1239 of corresponding 3 to 5 the example of communicating by letter. Go out not draw unit indicator (duration is 21 to 5 communication, and duration is 15 to 3 communication, is 43 to 5 communicate by letter with duration) on three reciprocations that wherein demonstrate.

Note always scheduled full until time=104 of keystone resources sender 3, be and reach the discriminant value that theoretic minimum scheduling time must reach. Should also be noted that recipient 1, inferior keystone resources is scheduled full until time=94. Although this logic not gross output is given birth to theoretic optimal scheduling table, it is designed to the resource with peak demand is effective.

If scheduling process is applied to the time period window of 100 units rather than the time period window of 150 units, request last in this time period window will be fallen by card from the request of having dispatched, but can be placed in the formation until the scheduling process of next time period window is restarted.

Top example has shown that the priority of all requests all is 1 required list. In one embodiment, support four different priority. According to logic, the request of the request of all highest levels what its priority level in office is scheduled before being considered and dispatching.

In case all guaranteed requests all have been scheduled, just be present in guaranteed request in the dispatch list around dispatch other three priority. When having existed in the dispatch list when connecting, computing time the section window size process with regard to more complicated, but logic is duplicate. Although ACR value (or SAV) is determining when to dispatch what time leading factor still the time, the connection request between two same parallel resources is selected according to priority. This means that the ACR logic may determine 4 to 5 communication example request of before 4 to 2 communication example request of a medium priority of a scheduling low priority of scheduling, but will not before 4 to 5 communication example request of a medium priority of scheduling, dispatch it.

Because the connection requirement of determining may occur in the specific moment rather than simply in certain time period window, just can adjust and make these requests get around scheduling. These solid times being connected between determining enter before the scheduling beginning event table and those connections that is scheduled must be scheduled in be fixedly connected with around. Connection between these solid times are definite has the highest priority in all requests significantly.

In above-mentioned example, front shift time is exactly such something, namely finds out simply the chronological next end of the connection of having dispatched. A lot of circulations but the process of this front shift time has been used up, and these circulations preferably are used for carrying out other function. The network dispatcher of this preferred embodiment use 256 units the minimum time section (this example have 25,000 chronomeres scheduling time the section window rather than the window of foregoing 150 chronomeres) to lower the demand of processing. Although this has obvious negative effect to dispatching efficiency, also there are not so significantly positive influences. Owing between the time period, handle more event, more resource for scheduling just arranged in each time period fifty-fifty. Along with more resources are competed simultaneously, those resources with higher key value of hitting seem the unlikely resource that other finishes already by some and substitute. The end product of this compromise is that computation cycles is saved, and bandwidth efficiency is obviously not influenced.

In above-mentioned example, all requests are by integrated scheduling-do not need interrupt requests. If any one request has longer than the time period window (being 150 chronomeres in above-mentioned example) of definition, in order to be scheduled, in any case it will be divided at least two parts. If can not avoid long message in the design, long low priority information is unlikely by decomposing this information and using lower priority indication to withstand the information of higher priority to the competition of same resource in this case.

In this preferred embodiment, it is optional until the piece of 1023 unit head's " part " to be broken down into " expiring " pieces of 1000 chronomeres and one greater than the request of 1023 chronomeres.

In this preferred embodiment, because the variation of route (being used for transmitting the specific information of dispatch list) and possible large capacity, they are the information of oneself's scheduling. But timetable (being used for collecting the information of connection request) is very little and be used as the connection of solid time between determining and process (they are not dispatched one by one). The scheduling flow that is used for this preferred embodiment is included in the collection of very first time section window timetable and the scheduling of circuit, wherein second, third and the 4th time period window (correspondingly be used for the scheduling of information, transmit the enforcement of dispatch list and connection) by according to as described in use.

Referring now to Figure 13,, it has described the representative corresponding to the parts of member node 107,108 in more detail. Although Figure 13 has continued to use the element number of Fig. 1 and Fig. 5, it has also shown the parallel sending and receiving that allows corresponding to the information of other node. In this case, the transmission logical one 301 and the RL 1302 that have shown master scheduling device 106, send logical one 319 and RL 1320 with following on member node 107, and on member node 108 another followed and sent logical one 321 and RL 1322. Path 1311,1305 and 1306 has shown from the information path of the outside vectorization of message carrier 100 (band crossbar switch 516). Use and to cross as discussed earlier flow process cycle calculations machine and just can obtain thisly making specific node 107,108,106 be busy with the highly communication means of the information sending and receiving of intersection. Node 107 is shown local scheduler 1307 by details drawing more, oscillator 1308, CPU1309 and VCXO (crystal oscillator control voltage) 1309, and the local task list 1303 of describing with the front and the local schedule of tasks 1304 of usually having been discussed show.

Referring now to focusing on three kinds of technical Figure 14,, in conjunction with it, the network about scattering has been described compendiously, the unique and valuable characteristic of the thinking of real-time computer is further to see characteristic of the present invention clearly comprehensively. From this point, the close synchronization of key technology and intercomputer; The operation that time is determined comprises network service; And the task to scattering, particularly those relate to central authorities' scheduling of the task of network service. This sketch has shown the correlation 1400 between the various technology of implementing in this preferred embodiment, has emphasized that wherein those are considered to for being implemented is crucial technology. These basic technology combine an architecture just are provided, and it can be responsible for the real-time calculating in the computing environment of scattering. These technology are (1) time definite event control 1422; (2) close synchronizationization between system 1423; (3) central authorities' scheduling 1421 of the task of the key of real-time system operation in the computing environment of scattering.

Relation between these technology is shown by arrow, wherein arrow is from allowing the realization technology to point to the realization technology that is allowed to, and wherein term " allows to realize " to use to be described in technology effectiveness aspect or the improvement aspect the simplification of this technology application here broadly. Notice that many technology are mutually useful, point out with four-headed arrow here.

Following table has been explained the relation of implementing between the characteristic, as Figure 14 in close synchronization 1423, priorization 1424, time definite event control 1422, safety 1426, limit examination 1425, central authoritiesization control 1421, reliability 1428, the virtual linkage between the dynamic redundancy 1427 and low-cost 1435.

The interaction of table 33 key technical feature

Connection among the figure	Effect
			1401	The central authoritiesization scheduling guarantees that the all-network reciprocation occurs between the rule time determines. In this was described, first demand that the time is determined was the node close synchronization that makes on the network. The neither one network node can be known the needs of whole network, thus do not having in the situation of central scheduler, the replacer or eliminate all dynamically, perhaps implement the scheme that can hold this variation. By central authoritiesization connection scheduling, the acting in conjunction of priorization and limit examination, interconnecting of (high priority) time-sensitive can be to determine the time, although interconnecting of (low priority) high power capacity may exist by the design that variation is held.
1402	In the computing environment of scattering, with the system of close synchronization, or be connected to realize that the bandwidth of enough affluences of in good time sending can obtain deterministic network connection by being provided at the worst calibration and bandwidth requirement situation. Because the form analysis of such worst case is very difficult, also may be very expensive because of the bandwidth that affluence is provided, synchronization determines to be exactly to allow the realization technology for the time. In addition, the difficulty when obtaining desired level synchronous is usually proportional with the uncertain degree of communicating by letter.
		1403	The time that dynamic redundancy mainly allows to restore about failure is determined. The possible failed situation of management is not difficult when having redundancy, yet the recovery of carrying out failure in the system that determines the time that positive redundancy arranged also is difficult. Time determines to allow to use dynamic redundancy, wherein to moving

	One of redundant maximum challenge of attitude just is to guarantee that each side moves samely. And then the environment of determining with a time just can not occured to move accordingly to diagnose by noticing in the regular hour when the unit of a redundancy of system generating function obstacle simply. The network that this time is determined comprises time-based communication between the computing unit of separation. The fact that communication is dispatched by time-based allows the use of dynamic redundancy. In dynamic redundancy, there are two or more computers sequentially to carry out same logic via the programmed instruction of oneself separately, exchange the result of this execution, then decide to provide the result. Because the data that each computer receives must be shared this fact, the time that must have certain mechanism to decide the ruling process to begin with other computer of carrying out same logic before deciding. This mechanism is chosen to be to be provided at and provides between redundant computer synchronously and according to the task of time scheduling about communication. Learn by this way the time that the ruling task can begin.
		1404	Challenge the most difficult implementing dynamic redundancy is the difference of processing between the each side. Must expect that in the checkpoint of determining the each side of redundant system is consistent with each other, and in order to keep this value, inconsistent must being eliminated. All each side must know when arrive the checkpoint, and the permission that must lay down a regulation is eliminated different. No matter the employing center-control is in terms of content in the connection request or the difference on time course can both be reduced even elimination, thereby increases the possibility that is equal to that all each side keep operation. On the other hand, the central authoritiesization scheduling greatly depends on the availability of scheduling unit. The central authoritiesization scheduling becomes more practical when having redundancy.
1405	The central schedule analysis that all connection requests need to be gathered together.

	This tolerance limit checks to prove that the combination of (1) certain undesired node-to-node is under an embargo and (2) traffic level of (may in specific priority) between certain node combination is no more than predeterminated level.
		1406	By guaranteeing not break through certain restriction, limit examination has strengthened the security of the computing system that scatters. These limit are checked in certain system and should be allowed to do the attack that the time spent helps opposing malice with other system interaction, and resist unexpected dispute, transship such as overload or the accidental of high-priority data of legal lower-priority data to system.
1407	The priorization of information has increased the application of limit examination. Can not only in the situation of restriction lower-priority data within reason, make high-priority data become that the time determines, and can it be placed by the information priority level.
		1408	By preventing security risk and definite by the retention time, limit examination has improved the overall reliability of system. If there is not limit examination, certain network error such as the accidental requested super traffic in the high priority level of a node, may damage network.
1409	Whether the examination limit priority traffic has violated the flux threshold value to guarantee not hinder the mistake that occurs in the internodal limit priority communication of any inerrancy at passage in the implementation process of limit examination.
		1410	The non-blocking network architecture allows network to realize the flux more taller than the possible flux of arbitrary specific node in the network. This means any specifically can with the cost compared of blocking-up network architecture under higher bandwidth is arranged.
1411	In the time can entrusting the TSM Security Agent Administrative Security policy of central authoritiesization, security just is easier to be implemented. Especially in this preferred embodiment, give security table of central scheduler, and mutual connected system can entrust to central manager with safe task, only allow in a place rather than this management is carried out in two places. Also have, repeat in each environment that same safety policy needn't exist in star graph.

1412	By introducing central scheduler, scheduling logic can be implemented in order to strengthen the priorization that has received connection request.
		1413	Definite task of enforcement time has been simplified in priorization. This is because large but not frequent task can be provided under the low priority and moves as program is written into, more simply but the analysis that height hits operation may be got rid of by more crucial peer on communication task at the same level.
1415	Using on the network of Radio Broadcasting Agreements, the all-network node all has an opportunity to see the data that other all nodes go (although encrypt may make this observation invalid). Not forever to implement the blocking-up agreement in Radio Broadcasting Agreements, unless clog-free agreement has been stopped the possibility of broadcasting, therefore avoided obvious security risk.
		1416	One of obvious method that obtains reliability is redundant introducing.
1417	By the introducing of priorization, the primary and foremost purpose of this high bandwidth of high flux is subject to positive influences in two ways. The first, those high-throughout perception with data of maximum volume are transmitted by calling in mode more timely of priorization. The second, when the data of the known high priority that those usually must in time be transmitted can not be influenced owing to the communication request that allows the large volume low priority, can allow the data of larger volume.
		1418	Dynamic redundancy needs failed unit to recover from fault time. This usually relates to from the data of trouble-free elements exchange record. For hitting of this exchange minimized, can introduce that priorization can recover with the unit of guaranteeing to recover and be not have the mode of actual influence to recover to hitting of healthy cells.
1419	The congenital risk that has in the shutdown of system-down network of some architecture and layout. Although using the conversion of clog-free architecture and physical circuit is not the unique method that obtains higher reliability, it provides the level of isolated network error.
		1420	For choke free, used certain crossbar switch such as those

	Effective utilization of the available bandwidth of network architecture, central schedule are one and allow the realization technology. If not scheduling just must be carried out fixing connection mode, therefore in fact reduced the flux of these nodes.
		1431	Owing to several reasons, it may be difficult that the practicality of physical circuit switch technology is got up, and this comprises and determines in officely to mean the fixed moment and should dispatch the suitable relevant challenge of what circuit. If know in advance needs, just can determine in advance the connection mode of circuit switching device and implement as required this pattern.
1432	Each system has an independent circuit or a pair of circuit just can implement to be scheduled to the circuit conversion of the system of tree topology in the network. The license of clog-free architecture just allows each passage to have its communication path.
		1433	Because the physical circuit transformation energy reduces the needs to dynamic circuit, high line speed usually causes the high network bandwidth.
1436	The circuit reduction that contacts with the physical circuit photograph allows to realize given performance with lower cost.

Referring now to Figure 15,, discussed corresponding to each member node 108,107 (control computers 506, logger computer 507, control computer 504, logger computer 505, user interface 503, any one in network support 502 and the network switch 501) in the general processing 1500 that is performed of scheduling. Each member node 108,107 is responsible for receiving common clock signal; With common clock signal synchronously; Master scheduling device 106 from network switch 501 receives local event table 114,116; Form the tasks carrying table and incorporate any attachment of a task of in local event table 114,116, not describing into this table according to the local event table 114,116 that receives; The execution list of executing the task; Generation communication request table and the master scheduling device 106 in network switch 501 send these communication request tables.

Step 1501 and 1502-are because network so highly depends on the time reference of an overall situation, and the first step 1501 on member node 108,107 is exactly to receive common clock signal. Common clock signal is used as the master scheduling device 106 of an information from network switch 501 and sends each member node 108,107 to. Member node 108,107 receives this clock signal in step 1501, enclose the time of actual reception, confirms that this information is effectively whole. The information that is identified is used as the input of synchronization process, and in step 1502, the expection time of reception of temporal information and the difference between the actual time of reception are determined (character of temporal information is a pulse rather than the information that bears date alternatively) alternatively.

Each member node 108,107 of step 1503 and 1504-now must with the common clock signal synchronization. This synchronized implementation can be broken down into two independent tasks: the first, and in step 1503, voltage-controlled oscillator (1308) is adjusted to the frequency that is complementary with the common clock signal source; The second, in step 1504, time calibration, the starting point to each time cycle that is defined by common clock signal occured at one time in member node 108,107 and the master scheduling device 106 in network switch 501.

In a single day step 1505-has finished the synchronization of the master scheduling device 106 of member node 108,107 in the network switch 501, and member node 108,107 must receive local event table 114,116 by the master scheduling device 106 from network switch 501. Local event table 114,116 is comprising the dispatch list of the communication that relates to this member node. The reception itself of noting local event table 114,116 is by master scheduling device 106 scheduling in the network switch 501. Therefore, the node that receives local event table 114,116 for the first time must all be available in any time that may receive local event table 114,116, to receive this information.

Next step is to form the tasks carrying table to step 1506-to member node 108,107. This tasks carrying table be by local event table 114,116 and member node 108,107 must carry out but the attachment of a task that local event table 114,116 does not have to describe form. The local event table 114,116 that master scheduling device 106 from network switch 501 transmits provides the time scheduling relevant with the sending and receiving task. In order to form the tasks carrying table, must carry out following substep:

Substep 1506A: the zero-time that the master scheduling device 106 of adjustment from network switch 501 receives is to count propagation delay. The time scheduling that master scheduling device 106 from network switch 501 receives is about the master scheduling device 106 in the network switch 501. Although master scheduling device 106 synchronizations in member node 108,107 and the network switch 501 still exist intrinsic to postpone in by the propagation of communication medium. Master scheduling device 106 in the network switch 501 must count this to postpone, or call duration time section window must be enough large so that this intrinsic can be postponed included as the overhead of the time of setting aside current connection. In this preferred embodiment, this delay is counted by member node.

Substep 1506B: dispatch the task that those depend on communication task. Some task must be dispatched (transmission of the data of for example sending) from memory before transmission; Also some task must be dispatched afterwards in reception (for example under the redundant computation environment ruling of the data between a plurality of different sources and transmit data to memory). Member node 108,107 is responsible for these tasks of scheduling.

Substep 1506C: dispatch those both not local event table 114,116 descriptions, also do not rely on the task of communication. These are tasks that member node 108,107 must be finished outside the task relevant with communication. These tasks may be for other I/O operation, localization demonstration etc.

Next step is the execution list of executing the task to step 1507-to member node 108,107. Each task in the table is comprising will being performed of task, the zero-time of task and any necessary parameter of this task of successfully finishing.

Step 1508-forms the communication request tables now to member node 108,107, and it is the table that member node 108,107 is wanted the member node 108,107 of communicating by letter with it. The information that this communication request table comprises is the source node designator, destination node designator and the total length of communicating by letter.

Step 1509-communication request table is transmitted to the master scheduling device 106 in the network switch 501 in order to consider it in its scheduling. At this moment, member node 108,107 has been finished all essential effects within this specific time cycle. Next step 1510 is to have determined whether to receive the next time with reference to time information. If so, so whole process just begins the repetition itself from the earliest the first step 1501. Do not receive the next time with reference to time information if also arrive, the logic of member node may enter step 1505 so.

As a deliberate embodiment, Figure 16 has represented the general data flow sketch 130 of Fig. 1 with data flow sketch 1600, wherein finishes communication in the mode of radio frequency, and finishes through RF transmitter and receiver and to send 1601,1602 and 1603.

By use be considered to special-purpose computer equipment the present invention is implemented, with the computer program code of the special function of computer and sensing process control (in other words these computers are in one embodiment, for example, software or logic, here logic refers to licensed Circuits System or gives the license software that Circuits System is carried out) combine establishment, when it is connected with a power supply with box lunch, just be energized and be activated to carry out its computer program code (software), this special-purpose computer is carried out about special function of the present invention under the computer program code guiding. This computer comprises electroactive parts that are called as CPU, the physical storage that at least one links to each other with CPU, an input keyboard that optionally links to each other with CPU, (it can include but not limited to printer, it is output as the printing terminal of the paper that prints for output display that optionally links to each other with CPU, cathode-ray tube display, and/or flat-panel monitor), computer clock impulse generator and one are the power supply line of all electroactive parts supplying energies of computer. And then central processing unit has a control bus and for temporarily holding data (for example Register or superimposer) or processing the special-purpose computer circuit of operation (for example add circuit) for carrying out master data; This special-purpose computer circuit links to each other with control bus by latch circuit in communication, and this latch circuit can be in conduction (communication) state or non-conduction (non-communication) state independently; Determine that in the special time set of all latch circuit states at quarter has determined the particular state of central processing unit; And when the independent instruction of computer program code was pointed to these latch circuits by the output of CPU connected reference and clock pulse generator, the state of central processing unit was correspondingly adjusted and is changed. This computer also has the computer program code of the form of commercial operation system alternatively, and this operating system works as the operating platform of computer program code, and the computer dependent program code that points to special function of the present invention can be performed. Put it briefly, the present invention in this embodiment with and be applied in the preferred embodiment by many computers, these computers have been programmed to offer technology before the practical application of the application target of carrying out appointment and said system; Each computer works as the media of practical function, provides and this function is function by computer program code. In certain embodiments, possibility electricity consumption, light or radiofrequency signal communication interface and communication medium are implemented communication linkage in the special-purpose computer of some extend type, these interfaces and media use transmitter and the receiver with electric energy jointly, and they link to each other with at least one CPU directly or indirectly.

Residing in computer program code in the computer provides at least one parts of this equipment. Computer program code may be the form of the interaction process of mutual link block, routine and subroutine, program object and/or installation, but is not limited to these forms. In some existing Computer Designs, modern special IC (ASIC) computer program code may partly be expressed by the hard-wired circuitry similar to traditional hard wire door gear that does not take clock execution simplification logical scheme at least. The other side of the present invention that the figure of this description has shown key modules in computer program code and a large amount of structural relations between the parts and the function of equipment of the present invention can be realized; In this, outside the combination of hardware and software, constructed the computing unit in the system.

More than in illustrated mode the present invention has been described. On this point, be significantly to those persons skilled in the art, in case provided the disclosed benefit in front, just may modify the specific embodiment of here describing now and do not deviate from spirit of the present invention. Such modification is considered to be included within the scope of the present invention, and this scope is only limited by the scope and spirit of the claim of appendix.

Claims (21)

1. one kind by computer-implemented method, and its allows, and a plurality of computer nodes comprise step with collaborative mode actuating logic in interactive data communication:

Alignment time between said a plurality of computer nodes is so that each computer node has substantially the same real-time time;

For each computer node is determined a communication scheduling table, above-mentioned dispatch list has been determined the different time for begin each corresponding communication in future;

Transmit its dispatch list to corresponding computer node;

Dispatch list by it begins communication at reasonable time from each node.

2. network architecture comprises:

Support the message carrier of multiple communication path;

Many computers that link to each other with above-mentioned message carrier, above-mentioned computer are programmed by by a plurality of different running statuses, comprise that the circulation of communications status comes definite task of time of implementation;

Can be by the synchronous clock system of maintenance in above-mentioned computer that is used for of above-mentioned computer access;

With the scheduler that above-mentioned message carrier links to each other, it is set up communication and determines communication sequence between the above-mentioned computer by causing task that above-mentioned message carrier was determined according to the above-mentioned time of being carried out by above-mentioned computer between selected computer.

3. one kind by computer-implemented method, it allow in a plurality of computer nodes a node with above-mentioned a plurality of computer nodes in the mode actuating logic of other collaborative computer, comprise step:

Message carrier is connected to each computer node in above-mentioned a plurality of computer node;

In a computer node, provide the master scheduling ability to set up the master scheduling device;

In each computer node, determine corresponding to the required communication request table of communicating by letter with each time other computer node of the above-mentioned computer node of a future time period;

Send above-mentioned communication request table via above-mentioned message carrier to above-mentioned master scheduling device from each computer node;

In each computer node via above-mentioned message carrier from above-mentioned master scheduling device received communication dispatch list;

In each computer node, from above-mentioned communication scheduling table, determine a local task list.

4. one kind by computer-implemented method, and it allows a plurality of member node with the cooperative mode actuating logic, comprises step:

Message carrier is connected to each member node in above-mentioned a plurality of member node;

A master scheduling device is provided in the data communication via above-mentioned message carrier and above-mentioned each member node;

For above-mentioned a plurality of member node are determined one with reference to moment time value;

Transmit above-mentioned with reference to moment time value via above-mentioned message carrier to above-mentioned each member node and above-mentioned master scheduling device;

In each above-mentioned member node, determine a communication request table be presented at above-mentioned member node in the future time period required with the communicating by letter of other member node;

Transmit above-mentioned communication request table via above-mentioned message carrier to above-mentioned master scheduling device from each member node;

Determine the communication of having dispatched between each member node of communication event time in above-mentioned master scheduling device, these times are corresponding to above-mentioned reference moment time value and each the above-mentioned communication table in each above-mentioned communication request table that receives;

In above-mentioned master scheduling device, be the definite communication scheduling table corresponding to above-mentioned communication of having dispatched of each member node;

Transmit above-mentioned communication scheduling table via above-mentioned message carrier to each corresponding member node from above-mentioned master scheduling device;

In above-mentioned each member node, determine the time difference between the time when time value sends above-mentioned member node to constantly in above-mentioned reference moment time value with above-mentioned reference;

In above-mentioned each member node, be that member node is determined local schedule of tasks by above-mentioned communication scheduling table and the above-mentioned time difference, so that this member node was communicated by letter with other member node via above-mentioned message carrier in the above-mentioned communication event time of having dispatched;

Carry out the communication of having dispatched between the above-mentioned member node corresponding to each modulated degree communication via above-mentioned message carrier according to above-mentioned local schedule of tasks.

5. method as claimed in claim 4, wherein above-mentioned definitely carried out by above-mentioned master scheduling device with reference to the step of time value constantly.

6. method as claimed in claim 4, wherein the above-mentioned step of determining the communication request table in each member node has limited the maximum quantity of above-mentioned communication in above-mentioned future time period.

7. method as claimed in claim 4, wherein each communication has duration attribute and above-mentioned duration property value less than a fixing maximum.

8. method as claimed in claim 4 is wherein above-mentionedly determined to be carried out by the system of length of a game with reference to the step of time value constantly for a plurality of member node.

9. one kind by computer-implemented method, and it allows a plurality of member node with the cooperative mode actuating logic, comprises step:

Message carrier is connected to each member node in above-mentioned a plurality of member node, and each node all has a CPU;

A master scheduling device is provided in the data communication via above-mentioned message carrier and above-mentioned each member node;

For above-mentioned a plurality of member node are determined a reference moment time value as the time in a moment;

Transmit above-mentioned with reference to moment time value via above-mentioned message carrier to above-mentioned each member node and above-mentioned master scheduling device;

The access to any other member node that above-mentioned member node is required is determined in a communication request table in each above-mentioned member node, to obtain the Future direction time period corresponding to a future time period;

Transmit above-mentioned communication request table via above-mentioned message carrier to above-mentioned master scheduling device from each member node;

In above-mentioned master scheduling device, determine corresponding to above-mentioned member node required to the communications event data object of each time access of any other member node with the acquisition Future direction time period, above-mentioned communications event data object has an access member identity property value, accessed member's identity property value and an access duration property value;

In above-mentioned master scheduling device, all communications event data objects of determining are collected in communication event series;

In above-mentioned master scheduling device for determining that in above-mentioned each member node in communication event series a total accessed value is as all have that all access the summation of duration property values in the communications event data object of accessed member node identity property value corresponding to above-mentioned member node;

All have the summation of all access duration property values in the communications event data object of accessing member node identity property value corresponding to above-mentioned member node in order to determine a total access value conduct in above-mentioned each member node in communication event series in above-mentioned master scheduling device;

In above-mentioned master scheduling device, will advance earliest available time corresponding to the above-mentioned future communications task scheduling between following access member node and accessed member node with reference to moment time value, this access member node has the highest total access value corresponding to above-mentioned communication event series, have the highest associated access duration property value corresponding to above-mentioned access node corresponding to this accessed member node of accessed member's identity property value in all communications event data objects in communication event series, the communication task in above-mentioned future has the task starting time that equates with earliest available time in the dispatch list, corresponding to the access member node identity property value of access member node with corresponding to the accessed member node identity property value of accessed member node;

In above-mentioned master scheduling device in the deletion future communications task of having dispatched corresponding to this accessed member node identity property in the communications event data object of accessed member node identity property and the future communications task of dispatching again those communications event data objects that not yet are scheduled are determined that communication event is serial;

In above-mentioned master scheduling device, repeat to determine total accessed value corresponding to the communication event series that redefines, determine total access value, scheduling future communications task, with all steps corresponding to the communications event data object of the member node of having dispatched of deletion, until all communications event data objects all are scheduled till the communication scheduling table of each member node;

Send above-mentioned communication scheduling table via above-mentioned message carrier to each corresponding member node from above-mentioned master scheduling device;

In each member node, determine the time difference between the time when time value sends above-mentioned member node to constantly in above-mentioned reference moment time value with above-mentioned reference;

In each member node, be that member node is determined local schedule of tasks so that this member node was communicated by letter with other member node via above-mentioned message carrier in the above-mentioned communication event time of having dispatched by above-mentioned communication scheduling table and the above-mentioned time difference;

Carry out corresponding to the communication of having dispatched between the member node of each modulated degree communication event via above-mentioned message carrier according to above-mentioned local schedule of tasks;

Communication task between the member object in a plurality of member objects that work as collective's group for scheduling by computer-implemented method, comprise the following steps:

A computer is provided;

In above-mentioned computer, access corresponding communications event data object to obtain a Future direction time period for each time to any other member object that each member object is definite and above-mentioned member object is required, above-mentioned communications event data object has access member identity property value, accessed member's identity property value and access duration property value;

In above-mentioned computer, all fixed communications event data objects are collected into a communication event series;

In above-mentioned computer for determining that at above-mentioned each member object in communication event series a total accessed value is as all access the summation of duration property values in the communications event data object that has accessed member object identity property value corresponding to all of above-mentioned member object;

All have the summation of all access duration property values in the communications event data object of accessing member object identity property value corresponding to above-mentioned member object in order to determine a total access value conduct at above-mentioned each member object in communication event series in above-mentioned computer;

Determine a series of continuous future time sections within a dispatching cycle corresponding to dispatch list;

In above-mentioned computer, based on first available communication event data object in above-mentioned communication event series, the future communications task scheduling to be advanced in the earliest available time in first time period, this communication occurs in

Have corresponding to the access member object of the highest above-mentioned access value of above-mentioned communication event series and

Have between the accessed member object corresponding to the highest above-mentioned accessed value of above-mentioned communication event series, if between above-mentioned accessed member object and above-mentioned access member object, there is the communications event data object,

The communication task in above-mentioned future has the task starting time that equates with earliest available time in the dispatch list, corresponding to access member's access member object object identity property value, corresponding to accessed member object identity property value and the above-mentioned access duration property value of accessed member object;

In above-mentioned computer from the scheduling candidate deletion in the future communications task of having dispatched corresponding to this accessed member object identity property in the communications event data object of accessed member object identity property and the future communications task of dispatching again those communications event data objects that not yet are scheduled are determined that communication event is serial;

In above-mentioned computer, repeat to determine total accessed value corresponding to the communication event series that redefines, determine total access value, scheduling future communications task, with all steps corresponding to the communications event data object of the member object of having dispatched of deletion, till communications event data object in above-mentioned first future time section all has been scheduled until all can be scheduled;

In above-mentioned computer, also again carry out definite total accessed value corresponding to the remainder of above-mentioned a plurality of future time sections corresponding to the communication event series that redefines thereupon, determine total access value, scheduling future communications task, with all steps corresponding to the communications event data object of the member object of having dispatched of deletion, and repeat these steps until all communications event data objects that can be scheduled in above-mentioned each future time section all have been scheduled, and all communications event data objects are scheduled all into till above-mentioned dispatching cycle.

11. such as the method for claim 10, wherein above-mentionedly determine to be limited to the Future direction time period required in cycle predetermined time, and above-mentioned scheduling is also corresponding to above-mentioned cycle predetermined time.

12. the method such as claim 11, and then comprise the step that message carrier is connected with each member object in above-mentioned a plurality of member objects, above-mentioned message carrier has a plurality of passages that allow above-mentioned communication event parallel, be used as a whole implementation corresponding to the above-mentioned scheduling step of above-mentioned a plurality of passages, wherein above-mentioned earliest available time is corresponding to any one passage in the above-mentioned message carrier.

13. such as the method for claim 12, wherein above-mentioned each member object is with the member node in the data communication of above-mentioned message carrier, each member node all has a CPU.

14. one kind for the member object scheduler task in a plurality of resource objects that work as collective's group by computer-implemented method, comprise the following steps:

A computer is provided;

In above-mentioned computer, determine a series of Future directions for above-mentioned collective group;

In above-mentioned computer, determine each resource object for each Future direction in a series of Future directions;

In above-mentioned computer, determine between any two resource objects each time reciprocation corresponding to above-mentioned a series of Future directions;

For above-mentioned each time reciprocation of acquisition is that above-mentioned each resource object is determined usage example, above-mentioned usage example has the duration of use property value in above-mentioned computer;

In above-mentioned computer, determine access resources object and accessed resource object corresponding to above-mentioned usage example;

At first, in above-mentioned computer to each access resources object determine a total access time value as all corresponding to using the summation of duration property values as all of the above-mentioned usage example of the resource object of access resources object;

Then, in above-mentioned computer to each access resources object determine a total accessed time value as all corresponding to using the summation of duration property values as all of the above-mentioned usage example of the resource object of accessed resource object;

In above-mentioned computer, determine a series of corresponding to the future time section in the dispatching cycle of dispatch list;

To advance based on first the Future reciprocation event scheduling that can execute the task in above-mentioned a series of Future directions the earliest available time in first future time section in above-mentioned computer, this reciprocation occurs in:

Have corresponding to the access resources object of the highest above-mentioned access value summation of above-mentioned a series of Future directions and

Have between the accessed resource object corresponding to the highest above-mentioned accessed value summation of above-mentioned a series of Future directions, if between above-mentioned accessed resource object and above-mentioned access resources object, have above-mentioned a series of Future direction,

Above-mentioned Future reciprocation event has the task starting time that equates with earliest available time in the dispatch list, corresponding to the access resources object of access resources object identity property value, corresponding to accessed resource object identity property value and the above-mentioned use duration property value of accessed resource object;

In above-mentioned computer from the scheduling candidate deletion in the Future reciprocation event of having dispatched corresponding to this accessed member object identity property in the usage example of accessed resource object and the Future reciprocation event of dispatching again those reciprocations that not yet are scheduled are determined that Future direction is serial;

In above-mentioned computer, repeat to determine total accessed value corresponding to the Future direction series that redefines, determine total access value, scheduling future communications task, with all steps corresponding to the usage example of the resource object of having dispatched of deletion, till usage example in above-mentioned first future time section all has been scheduled until all can be scheduled;

In above-mentioned computer, also again carry out definite total accessed value corresponding to each remainder of above-mentioned a plurality of future time sections corresponding to the Future direction series that redefines thereupon, determine total access value, scheduling Future reciprocation event, with all steps corresponding to the resource object of the resource object of having dispatched of deletion, and repeat these steps until all communications event data objects that can be scheduled in above-mentioned each future time section all have been scheduled, and all resource reciprocation events are scheduled all into till above-mentioned dispatching cycle.

15. a computer-implemented device, it allows, and a plurality of computer nodes comprise with collaborative mode actuating logic in interactive data communication:

The calibration real-time time is so that each computer node has the device of identical real-time time between above-mentioned a plurality of computer nodes;

Determine the device of communication scheduling table for each computer node, above-mentioned dispatch list has the different time that will begin each corresponding task in future;

Transmit the device of each dispatch list to corresponding computer node;

The device that begins each communication from each node in the time that its dispatch list is determined.

16. the computer-implemented device with message carrier comprises:

A plurality of member node, each member node are all being carried out data communication with above-mentioned message carrier, and each member node has

A CPU,

Be used for from above-mentioned message carrier receive with reference to the reference time information of time value constantly and be used for determining above-mentioned with reference to time value constantly with receive the time-phase displacement logic of the difference of above-mentioned reference time information,

Be used for being identified at each required internode communication of above-mentioned member node in the future time period and determine logic with the communication that is used for setting up the communication request table of communicating by letter corresponding to all above-mentioned internodes,

Be used for sending logic via above-mentioned message carrier to the communication request that above-mentioned master scheduling device transmits above-mentioned communication request table,

Be used for receiving the communication task RL of above-mentioned communication scheduling table via above-mentioned message carrier from above-mentioned master scheduling device,

Be used for determining from above-mentioned communication scheduling table and the above-mentioned time difference the integrated logic of task of local schedule of tasks,

Be used for implementing logic according to the communication that above-mentioned local schedule of tasks is communicated by letter with other member node via above-mentioned message carrier,

Via the master scheduling device in above-mentioned message carrier and the above-mentioned a plurality of member node data communication, above-mentioned master scheduling utensil has the scheduler logic of being carried out by CPU, and above-mentioned scheduler logic has

Be used for determining constantly to determine logic as the above-mentioned of a moment in the time with reference to the constantly reference of time value,

Be used for creating the reference time information with above-mentioned reference moment time value and transmit above-mentioned reference moment communication logic with reference to time information with being used for via above-mentioned message carrier,

Be used for via the communication request RL of above-mentioned message carrier from above-mentioned each the communication request table of above-mentioned member node reception,

Be used for dispatching corresponding to the above-mentioned communication event time with reference to time value constantly between the member node with the communication of communicating by letter corresponding to above-mentioned each internode of in the above-mentioned communication request table that respectively receives, listing and be used for further as determining that corresponding to each member node of above-mentioned communication of having dispatched the communication scheduling table of communication request table determines logic

Be used for via the member node communication scheduling response logic of above-mentioned message carrier to above-mentioned each the communication scheduling table of corresponding member node transmission.

17. such as the device of claim 16, wherein each member node also comprises a variable-ratio oscillator.

18. such as the device of claim 16, wherein each member node also comprises logic and oscillator rate adaptation logic time calibration.

19. such as the device of claim 16, wherein above-mentioned message carrier comprises a crossbar switch.

20. such as the device of claim 16, wherein above-mentioned message carrier comprises a network topology.

21. such as the device of claim 16, wherein above-mentioned message carrier comprises a RF communication system.

Images